4 |
* - Decoder Module - |
* - Decoder Module - |
5 |
* |
* |
6 |
* Copyright(C) 2002 MinChen <chenm001@163.com> |
* Copyright(C) 2002 MinChen <chenm001@163.com> |
7 |
* 2002-2004 Peter Ross <pross@xvid.org> |
* 2002-2010 Peter Ross <pross@xvid.org> |
8 |
* |
* |
9 |
* This program is free software ; you can redistribute it and/or modify |
* This program is free software ; you can redistribute it and/or modify |
10 |
* it under the terms of the GNU General Public License as published by |
* it under the terms of the GNU General Public License as published by |
20 |
* along with this program ; if not, write to the Free Software |
* along with this program ; if not, write to the Free Software |
21 |
* Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA |
* Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA |
22 |
* |
* |
23 |
* $Id: decoder.c,v 1.70 2005-05-17 21:03:32 Skal Exp $ |
* $Id$ |
24 |
* |
* |
25 |
****************************************************************************/ |
****************************************************************************/ |
26 |
|
|
61 |
#include "image/postprocessing.h" |
#include "image/postprocessing.h" |
62 |
#include "utils/mem_align.h" |
#include "utils/mem_align.h" |
63 |
|
|
64 |
|
#define DIV2ROUND(n) (((n)>>1)|((n)&1)) |
65 |
|
#define DIV2(n) ((n)>>1) |
66 |
|
#define DIVUVMOV(n) (((n) >> 1) + roundtab_79[(n) & 0x3]) // |
67 |
|
|
68 |
static int |
static int |
69 |
decoder_resize(DECODER * dec) |
decoder_resize(DECODER * dec) |
70 |
{ |
{ |
149 |
int |
int |
150 |
decoder_create(xvid_dec_create_t * create) |
decoder_create(xvid_dec_create_t * create) |
151 |
{ |
{ |
152 |
|
int ret = 0; |
153 |
DECODER *dec; |
DECODER *dec; |
154 |
|
|
155 |
if (XVID_VERSION_MAJOR(create->version) != 1) /* v1.x.x */ |
if (XVID_VERSION_MAJOR(create->version) != 1) /* v1.x.x */ |
170 |
|
|
171 |
create->handle = dec; |
create->handle = dec; |
172 |
|
|
173 |
dec->width = create->width; |
dec->width = MAX(0, create->width); |
174 |
dec->height = create->height; |
dec->height = MAX(0, create->height); |
175 |
|
|
176 |
|
dec->num_threads = MAX(0, create->num_threads); |
177 |
|
|
178 |
image_null(&dec->cur); |
image_null(&dec->cur); |
179 |
image_null(&dec->refn[0]); |
image_null(&dec->refn[0]); |
198 |
dec->low_delay = 0; |
dec->low_delay = 0; |
199 |
dec->packed_mode = 0; |
dec->packed_mode = 0; |
200 |
dec->time_inc_resolution = 1; /* until VOL header says otherwise */ |
dec->time_inc_resolution = 1; /* until VOL header says otherwise */ |
201 |
|
dec->ver_id = 1; |
202 |
|
|
203 |
|
if (create->fourcc == ((int)('X')|((int)('V')<<8)| |
204 |
|
((int)('I')<<16)|((int)('D')<<24))) { /* XVID */ |
205 |
|
dec->bs_version = 0; /* Initially assume oldest xvid version */ |
206 |
|
} |
207 |
|
else { |
208 |
|
dec->bs_version = 0xffff; /* Initialize to very high value -> assume bugfree stream */ |
209 |
|
} |
210 |
|
|
211 |
dec->fixed_dimensions = (dec->width > 0 && dec->height > 0); |
dec->fixed_dimensions = (dec->width > 0 && dec->height > 0); |
212 |
|
|
213 |
if (dec->fixed_dimensions) |
ret = decoder_resize(dec); |
214 |
return decoder_resize(dec); |
if (ret == XVID_ERR_MEMORY) create->handle = NULL; |
215 |
else |
|
216 |
return 0; |
return ret; |
217 |
} |
} |
218 |
|
|
219 |
|
|
264 |
uint32_t stride2 = stride / 2; |
uint32_t stride2 = stride / 2; |
265 |
uint32_t next_block = stride * 8; |
uint32_t next_block = stride * 8; |
266 |
uint32_t i; |
uint32_t i; |
267 |
uint32_t iQuant = pMB->quant; |
uint32_t iQuant = MAX(1, pMB->quant); |
268 |
uint8_t *pY_Cur, *pU_Cur, *pV_Cur; |
uint8_t *pY_Cur, *pU_Cur, *pV_Cur; |
269 |
|
|
270 |
pY_Cur = dec->cur.y + (y_pos << 4) * stride + (x_pos << 4); |
pY_Cur = dec->cur.y + (y_pos << 4) * stride + (x_pos << 4); |
328 |
stop_iquant_timer(); |
stop_iquant_timer(); |
329 |
|
|
330 |
start_timer(); |
start_timer(); |
331 |
idct(&data[i * 64]); |
idct((short * const)&data[i * 64]); |
332 |
stop_idct_timer(); |
stop_idct_timer(); |
333 |
|
|
334 |
} |
} |
360 |
DECLARE_ALIGNED_MATRIX(data, 1, 64, int16_t, CACHE_LINE); |
DECLARE_ALIGNED_MATRIX(data, 1, 64, int16_t, CACHE_LINE); |
361 |
|
|
362 |
int stride = dec->edged_width; |
int stride = dec->edged_width; |
|
int next_block = stride * 8; |
|
363 |
int i; |
int i; |
364 |
const uint32_t iQuant = pMB->quant; |
const uint32_t iQuant = MAX(1, pMB->quant); |
365 |
const int direction = dec->alternate_vertical_scan ? 2 : 0; |
const int direction = dec->alternate_vertical_scan ? 2 : 0; |
366 |
typedef void (*get_inter_block_function_t)( |
typedef void (*get_inter_block_function_t)( |
367 |
Bitstream * bs, |
Bitstream * bs, |
383 |
|
|
384 |
|
|
385 |
if (dec->interlacing && pMB->field_dct) { |
if (dec->interlacing && pMB->field_dct) { |
|
next_block = stride; |
|
|
stride *= 2; |
|
|
} |
|
|
|
|
386 |
dst[0] = pY_Cur; |
dst[0] = pY_Cur; |
387 |
dst[2] = pY_Cur + next_block; |
dst[1] = pY_Cur + 8; |
388 |
dst[1] = dst[0] + 8; |
dst[2] = pY_Cur + stride; |
389 |
|
dst[3] = dst[2] + 8; |
390 |
|
dst[4] = pU_Cur; |
391 |
|
dst[5] = pV_Cur; |
392 |
|
strides[0] = strides[1] = strides[2] = strides[3] = stride*2; |
393 |
|
strides[4] = stride/2; |
394 |
|
strides[5] = stride/2; |
395 |
|
} else { |
396 |
|
dst[0] = pY_Cur; |
397 |
|
dst[1] = pY_Cur + 8; |
398 |
|
dst[2] = pY_Cur + 8*stride; |
399 |
dst[3] = dst[2] + 8; |
dst[3] = dst[2] + 8; |
400 |
dst[4] = pU_Cur; |
dst[4] = pU_Cur; |
401 |
dst[5] = pV_Cur; |
dst[5] = pV_Cur; |
402 |
strides[0] = strides[1] = strides[2] = strides[3] = stride; |
strides[0] = strides[1] = strides[2] = strides[3] = stride; |
403 |
strides[4] = stride/2; |
strides[4] = stride/2; |
404 |
strides[5] = stride/2; |
strides[5] = stride/2; |
405 |
|
} |
406 |
|
|
407 |
for (i = 0; i < 6; i++) { |
for (i = 0; i < 6; i++) { |
408 |
/* Process only coded blocks */ |
/* Process only coded blocks */ |
418 |
|
|
419 |
/* iDCT */ |
/* iDCT */ |
420 |
start_timer(); |
start_timer(); |
421 |
idct(&data[0]); |
idct((short * const)&data[0]); |
422 |
stop_idct_timer(); |
stop_idct_timer(); |
423 |
|
|
424 |
/* Add this residual to the predicted block */ |
/* Add this residual to the predicted block */ |
465 |
CHECK_MV(mv[3]); |
CHECK_MV(mv[3]); |
466 |
} |
} |
467 |
|
|
468 |
|
/* Up to this version, chroma rounding was wrong with qpel. |
469 |
|
* So we try to be backward compatible to avoid artifacts */ |
470 |
|
#define BS_VERSION_BUGGY_CHROMA_ROUNDING 1 |
471 |
|
|
472 |
/* decode an inter macroblock */ |
/* decode an inter macroblock */ |
473 |
static void |
static void |
474 |
decoder_mbinter(DECODER * dec, |
decoder_mbinter(DECODER * dec, |
478 |
const uint32_t cbp, |
const uint32_t cbp, |
479 |
Bitstream * bs, |
Bitstream * bs, |
480 |
const uint32_t rounding, |
const uint32_t rounding, |
481 |
const int ref) |
const int ref, |
482 |
|
const int bvop) |
483 |
{ |
{ |
484 |
uint32_t stride = dec->edged_width; |
uint32_t stride = dec->edged_width; |
485 |
uint32_t stride2 = stride / 2; |
uint32_t stride2 = stride / 2; |
500 |
|
|
501 |
start_timer(); |
start_timer(); |
502 |
|
|
503 |
if (pMB->mode != MODE_INTER4V) { /* INTER, INTER_Q, NOT_CODED, FORWARD, BACKWARD */ |
if ((pMB->mode != MODE_INTER4V) || (bvop)) { /* INTER, INTER_Q, NOT_CODED, FORWARD, BACKWARD */ |
504 |
|
|
505 |
uv_dx = mv[0].x; |
uv_dx = mv[0].x; |
506 |
uv_dy = mv[0].y; |
uv_dy = mv[0].y; |
507 |
if (dec->quarterpel) { |
if (dec->quarterpel) { |
508 |
|
if (dec->bs_version <= BS_VERSION_BUGGY_CHROMA_ROUNDING) { |
509 |
|
uv_dx = (uv_dx>>1) | (uv_dx&1); |
510 |
|
uv_dy = (uv_dy>>1) | (uv_dy&1); |
511 |
|
} |
512 |
|
else { |
513 |
uv_dx /= 2; |
uv_dx /= 2; |
514 |
uv_dy /= 2; |
uv_dy /= 2; |
515 |
} |
} |
516 |
|
} |
517 |
uv_dx = (uv_dx >> 1) + roundtab_79[uv_dx & 0x3]; |
uv_dx = (uv_dx >> 1) + roundtab_79[uv_dx & 0x3]; |
518 |
uv_dy = (uv_dy >> 1) + roundtab_79[uv_dy & 0x3]; |
uv_dy = (uv_dy >> 1) + roundtab_79[uv_dy & 0x3]; |
519 |
|
|
528 |
} else { /* MODE_INTER4V */ |
} else { /* MODE_INTER4V */ |
529 |
|
|
530 |
if(dec->quarterpel) { |
if(dec->quarterpel) { |
531 |
|
if (dec->bs_version <= BS_VERSION_BUGGY_CHROMA_ROUNDING) { |
532 |
|
int z; |
533 |
|
uv_dx = 0; uv_dy = 0; |
534 |
|
for (z = 0; z < 4; z++) { |
535 |
|
uv_dx += ((mv[z].x>>1) | (mv[z].x&1)); |
536 |
|
uv_dy += ((mv[z].y>>1) | (mv[z].y&1)); |
537 |
|
} |
538 |
|
} |
539 |
|
else { |
540 |
uv_dx = (mv[0].x / 2) + (mv[1].x / 2) + (mv[2].x / 2) + (mv[3].x / 2); |
uv_dx = (mv[0].x / 2) + (mv[1].x / 2) + (mv[2].x / 2) + (mv[3].x / 2); |
541 |
uv_dy = (mv[0].y / 2) + (mv[1].y / 2) + (mv[2].y / 2) + (mv[3].y / 2); |
uv_dy = (mv[0].y / 2) + (mv[1].y / 2) + (mv[2].y / 2) + (mv[3].y / 2); |
542 |
|
} |
543 |
} else { |
} else { |
544 |
uv_dx = mv[0].x + mv[1].x + mv[2].x + mv[3].x; |
uv_dx = mv[0].x + mv[1].x + mv[2].x + mv[3].x; |
545 |
uv_dy = mv[0].y + mv[1].y + mv[2].y + mv[3].y; |
uv_dy = mv[0].y + mv[1].y + mv[2].y + mv[3].y; |
585 |
decoder_mb_decode(dec, cbp, bs, pY_Cur, pU_Cur, pV_Cur, pMB); |
decoder_mb_decode(dec, cbp, bs, pY_Cur, pU_Cur, pV_Cur, pMB); |
586 |
} |
} |
587 |
|
|
588 |
|
/* decode an inter macroblock in field mode */ |
589 |
|
static void |
590 |
|
decoder_mbinter_field(DECODER * dec, |
591 |
|
const MACROBLOCK * pMB, |
592 |
|
const uint32_t x_pos, |
593 |
|
const uint32_t y_pos, |
594 |
|
const uint32_t cbp, |
595 |
|
Bitstream * bs, |
596 |
|
const uint32_t rounding, |
597 |
|
const int ref, |
598 |
|
const int bvop) |
599 |
|
{ |
600 |
|
uint32_t stride = dec->edged_width; |
601 |
|
uint32_t stride2 = stride / 2; |
602 |
|
|
603 |
|
uint8_t *pY_Cur, *pU_Cur, *pV_Cur; |
604 |
|
|
605 |
|
int uvtop_dx, uvtop_dy; |
606 |
|
int uvbot_dx, uvbot_dy; |
607 |
|
VECTOR mv[4]; /* local copy of mvs */ |
608 |
|
|
609 |
|
/* Get pointer to memory areas */ |
610 |
|
pY_Cur = dec->cur.y + (y_pos << 4) * stride + (x_pos << 4); |
611 |
|
pU_Cur = dec->cur.u + (y_pos << 3) * stride2 + (x_pos << 3); |
612 |
|
pV_Cur = dec->cur.v + (y_pos << 3) * stride2 + (x_pos << 3); |
613 |
|
|
614 |
|
mv[0] = pMB->mvs[0]; |
615 |
|
mv[1] = pMB->mvs[1]; |
616 |
|
memset(&mv[2],0,2*sizeof(VECTOR)); |
617 |
|
|
618 |
|
validate_vector(mv, x_pos, y_pos, dec); |
619 |
|
|
620 |
|
start_timer(); |
621 |
|
|
622 |
|
if((pMB->mode!=MODE_INTER4V) || (bvop)) /* INTER, INTER_Q, NOT_CODED, FORWARD, BACKWARD */ |
623 |
|
{ |
624 |
|
/* Prepare top field vector */ |
625 |
|
uvtop_dx = DIV2ROUND(mv[0].x); |
626 |
|
uvtop_dy = DIV2ROUND(mv[0].y); |
627 |
|
|
628 |
|
/* Prepare bottom field vector */ |
629 |
|
uvbot_dx = DIV2ROUND(mv[1].x); |
630 |
|
uvbot_dy = DIV2ROUND(mv[1].y); |
631 |
|
|
632 |
|
if(dec->quarterpel) |
633 |
|
{ |
634 |
|
/* NOT supported */ |
635 |
|
} |
636 |
|
else |
637 |
|
{ |
638 |
|
/* Interpolate top field left part(we use double stride for every 2nd line) */ |
639 |
|
interpolate8x8_switch(dec->cur.y,dec->refn[ref].y+pMB->field_for_top*stride, |
640 |
|
16*x_pos,8*y_pos,mv[0].x, mv[0].y>>1,2*stride, rounding); |
641 |
|
/* top field right part */ |
642 |
|
interpolate8x8_switch(dec->cur.y,dec->refn[ref].y+pMB->field_for_top*stride, |
643 |
|
16*x_pos+8,8*y_pos,mv[0].x, mv[0].y>>1,2*stride, rounding); |
644 |
|
|
645 |
|
/* Interpolate bottom field left part(we use double stride for every 2nd line) */ |
646 |
|
interpolate8x8_switch(dec->cur.y+stride,dec->refn[ref].y+pMB->field_for_bot*stride, |
647 |
|
16*x_pos,8*y_pos,mv[1].x, mv[1].y>>1,2*stride, rounding); |
648 |
|
/* Bottom field right part */ |
649 |
|
interpolate8x8_switch(dec->cur.y+stride,dec->refn[ref].y+pMB->field_for_bot*stride, |
650 |
|
16*x_pos+8,8*y_pos,mv[1].x, mv[1].y>>1,2*stride, rounding); |
651 |
|
|
652 |
|
/* Interpolate field1 U */ |
653 |
|
interpolate8x4_switch(dec->cur.u,dec->refn[ref].u+pMB->field_for_top*stride2, |
654 |
|
8*x_pos,4*y_pos,uvtop_dx,DIV2ROUND(uvtop_dy),stride,rounding); |
655 |
|
|
656 |
|
/* Interpolate field1 V */ |
657 |
|
interpolate8x4_switch(dec->cur.v,dec->refn[ref].v+pMB->field_for_top*stride2, |
658 |
|
8*x_pos,4*y_pos,uvtop_dx,DIV2ROUND(uvtop_dy),stride,rounding); |
659 |
|
|
660 |
|
/* Interpolate field2 U */ |
661 |
|
interpolate8x4_switch(dec->cur.u+stride2,dec->refn[ref].u+pMB->field_for_bot*stride2, |
662 |
|
8*x_pos,4*y_pos,uvbot_dx,DIV2ROUND(uvbot_dy),stride,rounding); |
663 |
|
|
664 |
|
/* Interpolate field2 V */ |
665 |
|
interpolate8x4_switch(dec->cur.v+stride2,dec->refn[ref].v+pMB->field_for_bot*stride2, |
666 |
|
8*x_pos,4*y_pos,uvbot_dx,DIV2ROUND(uvbot_dy),stride,rounding); |
667 |
|
} |
668 |
|
} |
669 |
|
else |
670 |
|
{ |
671 |
|
/* We don't expect 4 motion vectors in interlaced mode */ |
672 |
|
} |
673 |
|
|
674 |
|
stop_comp_timer(); |
675 |
|
|
676 |
|
/* Must add error correction? */ |
677 |
|
if(cbp) |
678 |
|
decoder_mb_decode(dec, cbp, bs, pY_Cur, pU_Cur, pV_Cur, pMB); |
679 |
|
} |
680 |
|
|
681 |
static void |
static void |
682 |
decoder_mbgmc(DECODER * dec, |
decoder_mbgmc(DECODER * dec, |
683 |
MACROBLOCK * const pMB, |
MACROBLOCK * const pMB, |
757 |
bound = read_video_packet_header(bs, dec, 0, |
bound = read_video_packet_header(bs, dec, 0, |
758 |
&quant, NULL, NULL, &intra_dc_threshold); |
&quant, NULL, NULL, &intra_dc_threshold); |
759 |
x = bound % mb_width; |
x = bound % mb_width; |
760 |
y = bound / mb_width; |
y = MIN((bound / mb_width), (mb_height-1)); |
761 |
} |
} |
762 |
mb = &dec->mbs[y * dec->mb_width + x]; |
mb = &dec->mbs[y * dec->mb_width + x]; |
763 |
|
|
845 |
ret_mv->y = mv.y; |
ret_mv->y = mv.y; |
846 |
} |
} |
847 |
|
|
848 |
|
/* We use this when decoder runs interlaced -> different prediction */ |
849 |
|
|
850 |
|
static void get_motion_vector_interlaced(DECODER * dec, |
851 |
|
Bitstream * bs, |
852 |
|
int x, |
853 |
|
int y, |
854 |
|
int k, |
855 |
|
MACROBLOCK *pMB, |
856 |
|
int fcode, |
857 |
|
const int bound) |
858 |
|
{ |
859 |
|
const int scale_fac = 1 << (fcode - 1); |
860 |
|
const int high = (32 * scale_fac) - 1; |
861 |
|
const int low = ((-32) * scale_fac); |
862 |
|
const int range = (64 * scale_fac); |
863 |
|
|
864 |
|
/* Get interlaced prediction */ |
865 |
|
const VECTOR pmv=get_pmv2_interlaced(dec->mbs,dec->mb_width,bound,x,y,k); |
866 |
|
VECTOR mv,mvf1,mvf2; |
867 |
|
|
868 |
|
if(!pMB->field_pred) |
869 |
|
{ |
870 |
|
mv.x = get_mv(bs,fcode); |
871 |
|
mv.y = get_mv(bs,fcode); |
872 |
|
|
873 |
|
mv.x += pmv.x; |
874 |
|
mv.y += pmv.y; |
875 |
|
|
876 |
|
if(mv.x<low) { |
877 |
|
mv.x += range; |
878 |
|
} else if (mv.x>high) { |
879 |
|
mv.x-=range; |
880 |
|
} |
881 |
|
|
882 |
|
if (mv.y < low) { |
883 |
|
mv.y += range; |
884 |
|
} else if (mv.y > high) { |
885 |
|
mv.y -= range; |
886 |
|
} |
887 |
|
|
888 |
|
pMB->mvs[0]=pMB->mvs[1]=pMB->mvs[2]=pMB->mvs[3]=mv; |
889 |
|
} |
890 |
|
else |
891 |
|
{ |
892 |
|
mvf1.x = get_mv(bs, fcode); |
893 |
|
mvf1.y = get_mv(bs, fcode); |
894 |
|
|
895 |
|
mvf1.x += pmv.x; |
896 |
|
mvf1.y = 2*(mvf1.y+pmv.y/2); /* It's multiple of 2 */ |
897 |
|
|
898 |
|
if (mvf1.x < low) { |
899 |
|
mvf1.x += range; |
900 |
|
} else if (mvf1.x > high) { |
901 |
|
mvf1.x -= range; |
902 |
|
} |
903 |
|
|
904 |
|
if (mvf1.y < low) { |
905 |
|
mvf1.y += range; |
906 |
|
} else if (mvf1.y > high) { |
907 |
|
mvf1.y -= range; |
908 |
|
} |
909 |
|
|
910 |
|
mvf2.x = get_mv(bs, fcode); |
911 |
|
mvf2.y = get_mv(bs, fcode); |
912 |
|
|
913 |
|
mvf2.x += pmv.x; |
914 |
|
mvf2.y = 2*(mvf2.y+pmv.y/2); /* It's multiple of 2 */ |
915 |
|
|
916 |
|
if (mvf2.x < low) { |
917 |
|
mvf2.x += range; |
918 |
|
} else if (mvf2.x > high) { |
919 |
|
mvf2.x -= range; |
920 |
|
} |
921 |
|
|
922 |
|
if (mvf2.y < low) { |
923 |
|
mvf2.y += range; |
924 |
|
} else if (mvf2.y > high) { |
925 |
|
mvf2.y -= range; |
926 |
|
} |
927 |
|
|
928 |
|
pMB->mvs[0]=mvf1; |
929 |
|
pMB->mvs[1]=mvf2; |
930 |
|
pMB->mvs[2].x=pMB->mvs[3].x=0; |
931 |
|
pMB->mvs[2].y=pMB->mvs[3].y=0; |
932 |
|
|
933 |
|
/* Calculate average for as it is field predicted */ |
934 |
|
pMB->mvs_avg.x=DIV2ROUND(pMB->mvs[0].x+pMB->mvs[1].x); |
935 |
|
pMB->mvs_avg.y=DIV2ROUND(pMB->mvs[0].y+pMB->mvs[1].y); |
936 |
|
} |
937 |
|
} |
938 |
|
|
939 |
/* for P_VOP set gmc_warp to NULL */ |
/* for P_VOP set gmc_warp to NULL */ |
940 |
static void |
static void |
941 |
decoder_pframe(DECODER * dec, |
decoder_pframe(DECODER * dec, |
984 |
bound = read_video_packet_header(bs, dec, fcode - 1, |
bound = read_video_packet_header(bs, dec, fcode - 1, |
985 |
&quant, &fcode, NULL, &intra_dc_threshold); |
&quant, &fcode, NULL, &intra_dc_threshold); |
986 |
x = bound % mb_width; |
x = bound % mb_width; |
987 |
y = bound / mb_width; |
y = MIN((bound / mb_width), (mb_height-1)); |
988 |
} |
} |
989 |
mb = &dec->mbs[y * dec->mb_width + x]; |
mb = &dec->mbs[y * dec->mb_width + x]; |
990 |
|
|
1028 |
} |
} |
1029 |
mb->quant = quant; |
mb->quant = quant; |
1030 |
|
|
1031 |
|
mb->field_pred=0; |
1032 |
if (dec->interlacing) { |
if (dec->interlacing) { |
1033 |
if (cbp || intra) { |
if (cbp || intra) { |
1034 |
mb->field_dct = BitstreamGetBit(bs); |
mb->field_dct = BitstreamGetBit(bs); |
1054 |
|
|
1055 |
} else if (mb->mode == MODE_INTER || mb->mode == MODE_INTER_Q) { |
} else if (mb->mode == MODE_INTER || mb->mode == MODE_INTER_Q) { |
1056 |
|
|
1057 |
if (dec->interlacing && mb->field_pred) { |
if(dec->interlacing) { |
1058 |
get_motion_vector(dec, bs, x, y, 0, &mb->mvs[0], fcode, bound); |
/* Get motion vectors interlaced, field_pred is handled there */ |
1059 |
get_motion_vector(dec, bs, x, y, 0, &mb->mvs[1], fcode, bound); |
get_motion_vector_interlaced(dec, bs, x, y, 0, mb, fcode, bound); |
1060 |
} else { |
} else { |
1061 |
get_motion_vector(dec, bs, x, y, 0, &mb->mvs[0], fcode, bound); |
get_motion_vector(dec, bs, x, y, 0, &mb->mvs[0], fcode, bound); |
1062 |
mb->mvs[1] = mb->mvs[2] = mb->mvs[3] = mb->mvs[0]; |
mb->mvs[1] = mb->mvs[2] = mb->mvs[3] = mb->mvs[0]; |
1063 |
} |
} |
1064 |
} else if (mb->mode == MODE_INTER4V ) { |
} else if (mb->mode == MODE_INTER4V ) { |
1065 |
|
/* interlaced missing here */ |
1066 |
get_motion_vector(dec, bs, x, y, 0, &mb->mvs[0], fcode, bound); |
get_motion_vector(dec, bs, x, y, 0, &mb->mvs[0], fcode, bound); |
1067 |
get_motion_vector(dec, bs, x, y, 1, &mb->mvs[1], fcode, bound); |
get_motion_vector(dec, bs, x, y, 1, &mb->mvs[1], fcode, bound); |
1068 |
get_motion_vector(dec, bs, x, y, 2, &mb->mvs[2], fcode, bound); |
get_motion_vector(dec, bs, x, y, 2, &mb->mvs[2], fcode, bound); |
1075 |
continue; |
continue; |
1076 |
} |
} |
1077 |
|
|
1078 |
decoder_mbinter(dec, mb, x, y, cbp, bs, rounding, 0); |
/* See how to decode */ |
1079 |
|
if(!mb->field_pred) |
1080 |
|
decoder_mbinter(dec, mb, x, y, cbp, bs, rounding, 0, 0); |
1081 |
|
else |
1082 |
|
decoder_mbinter_field(dec, mb, x, y, cbp, bs, rounding, 0, 0); |
1083 |
|
|
1084 |
} else if (gmc_warp) { /* a not coded S(GMC)-VOP macroblock */ |
} else if (gmc_warp) { /* a not coded S(GMC)-VOP macroblock */ |
1085 |
mb->mode = MODE_NOT_CODED_GMC; |
mb->mode = MODE_NOT_CODED_GMC; |
1097 |
|
|
1098 |
mb->mvs[0].x = mb->mvs[1].x = mb->mvs[2].x = mb->mvs[3].x = 0; |
mb->mvs[0].x = mb->mvs[1].x = mb->mvs[2].x = mb->mvs[3].x = 0; |
1099 |
mb->mvs[0].y = mb->mvs[1].y = mb->mvs[2].y = mb->mvs[3].y = 0; |
mb->mvs[0].y = mb->mvs[1].y = mb->mvs[2].y = mb->mvs[3].y = 0; |
1100 |
|
mb->field_pred=0; /* (!) */ |
1101 |
|
|
1102 |
decoder_mbinter(dec, mb, x, y, 0, bs, |
decoder_mbinter(dec, mb, x, y, 0, bs, |
1103 |
rounding, 0); |
rounding, 0, 0); |
1104 |
|
|
1105 |
if(dec->out_frm && cp_mb > 0) { |
if(dec->out_frm && cp_mb > 0) { |
1106 |
output_slice(&dec->cur, dec->edged_width,dec->width,dec->out_frm,st_mb,y,cp_mb); |
output_slice(&dec->cur, dec->edged_width,dec->width,dec->out_frm,st_mb,y,cp_mb); |
1182 |
b_uv_dy = pMB->b_mvs[0].y; |
b_uv_dy = pMB->b_mvs[0].y; |
1183 |
|
|
1184 |
if (dec->quarterpel) { |
if (dec->quarterpel) { |
1185 |
|
if (dec->bs_version <= BS_VERSION_BUGGY_CHROMA_ROUNDING) { |
1186 |
|
uv_dx = (uv_dx>>1) | (uv_dx&1); |
1187 |
|
uv_dy = (uv_dy>>1) | (uv_dy&1); |
1188 |
|
b_uv_dx = (b_uv_dx>>1) | (b_uv_dx&1); |
1189 |
|
b_uv_dy = (b_uv_dy>>1) | (b_uv_dy&1); |
1190 |
|
} |
1191 |
|
else { |
1192 |
uv_dx /= 2; |
uv_dx /= 2; |
1193 |
uv_dy /= 2; |
uv_dy /= 2; |
1194 |
b_uv_dx /= 2; |
b_uv_dx /= 2; |
1195 |
b_uv_dy /= 2; |
b_uv_dy /= 2; |
1196 |
} |
} |
1197 |
|
} |
1198 |
|
|
1199 |
uv_dx = (uv_dx >> 1) + roundtab_79[uv_dx & 0x3]; |
uv_dx = (uv_dx >> 1) + roundtab_79[uv_dx & 0x3]; |
1200 |
uv_dy = (uv_dy >> 1) + roundtab_79[uv_dy & 0x3]; |
uv_dy = (uv_dy >> 1) + roundtab_79[uv_dy & 0x3]; |
1202 |
b_uv_dy = (b_uv_dy >> 1) + roundtab_79[b_uv_dy & 0x3]; |
b_uv_dy = (b_uv_dy >> 1) + roundtab_79[b_uv_dy & 0x3]; |
1203 |
|
|
1204 |
} else { |
} else { |
1205 |
|
if (dec->quarterpel) { /* for qpel the /2 shall be done before summation. We've done it right in the encoder in the past. */ |
1206 |
|
/* TODO: figure out if we ever did it wrong on the encoder side. If yes, add some workaround */ |
1207 |
|
if (dec->bs_version <= BS_VERSION_BUGGY_CHROMA_ROUNDING) { |
1208 |
|
int z; |
1209 |
|
uv_dx = 0; uv_dy = 0; |
1210 |
|
b_uv_dx = 0; b_uv_dy = 0; |
1211 |
|
for (z = 0; z < 4; z++) { |
1212 |
|
uv_dx += ((pMB->mvs[z].x>>1) | (pMB->mvs[z].x&1)); |
1213 |
|
uv_dy += ((pMB->mvs[z].y>>1) | (pMB->mvs[z].y&1)); |
1214 |
|
b_uv_dx += ((pMB->b_mvs[z].x>>1) | (pMB->b_mvs[z].x&1)); |
1215 |
|
b_uv_dy += ((pMB->b_mvs[z].y>>1) | (pMB->b_mvs[z].y&1)); |
1216 |
|
} |
1217 |
|
} |
1218 |
|
else { |
1219 |
|
uv_dx = (pMB->mvs[0].x / 2) + (pMB->mvs[1].x / 2) + (pMB->mvs[2].x / 2) + (pMB->mvs[3].x / 2); |
1220 |
|
uv_dy = (pMB->mvs[0].y / 2) + (pMB->mvs[1].y / 2) + (pMB->mvs[2].y / 2) + (pMB->mvs[3].y / 2); |
1221 |
|
b_uv_dx = (pMB->b_mvs[0].x / 2) + (pMB->b_mvs[1].x / 2) + (pMB->b_mvs[2].x / 2) + (pMB->b_mvs[3].x / 2); |
1222 |
|
b_uv_dy = (pMB->b_mvs[0].y / 2) + (pMB->b_mvs[1].y / 2) + (pMB->b_mvs[2].y / 2) + (pMB->b_mvs[3].y / 2); |
1223 |
|
} |
1224 |
|
} else { |
1225 |
uv_dx = pMB->mvs[0].x + pMB->mvs[1].x + pMB->mvs[2].x + pMB->mvs[3].x; |
uv_dx = pMB->mvs[0].x + pMB->mvs[1].x + pMB->mvs[2].x + pMB->mvs[3].x; |
1226 |
uv_dy = pMB->mvs[0].y + pMB->mvs[1].y + pMB->mvs[2].y + pMB->mvs[3].y; |
uv_dy = pMB->mvs[0].y + pMB->mvs[1].y + pMB->mvs[2].y + pMB->mvs[3].y; |
1227 |
b_uv_dx = pMB->b_mvs[0].x + pMB->b_mvs[1].x + pMB->b_mvs[2].x + pMB->b_mvs[3].x; |
b_uv_dx = pMB->b_mvs[0].x + pMB->b_mvs[1].x + pMB->b_mvs[2].x + pMB->b_mvs[3].x; |
1228 |
b_uv_dy = pMB->b_mvs[0].y + pMB->b_mvs[1].y + pMB->b_mvs[2].y + pMB->b_mvs[3].y; |
b_uv_dy = pMB->b_mvs[0].y + pMB->b_mvs[1].y + pMB->b_mvs[2].y + pMB->b_mvs[3].y; |
|
|
|
|
if (dec->quarterpel) { |
|
|
uv_dx /= 2; |
|
|
uv_dy /= 2; |
|
|
b_uv_dx /= 2; |
|
|
b_uv_dy /= 2; |
|
1229 |
} |
} |
1230 |
|
|
1231 |
uv_dx = (uv_dx >> 3) + roundtab_76[uv_dx & 0xf]; |
uv_dx = (uv_dx >> 3) + roundtab_76[uv_dx & 0xf]; |
1344 |
return -1; |
return -1; |
1345 |
} |
} |
1346 |
|
|
1347 |
|
static int __inline get_resync_len_b(const int fcode_backward, |
1348 |
|
const int fcode_forward) { |
1349 |
|
int resync_len = ((fcode_forward>fcode_backward) ? fcode_forward : fcode_backward) - 1; |
1350 |
|
if (resync_len < 1) resync_len = 1; |
1351 |
|
return resync_len; |
1352 |
|
} |
1353 |
|
|
1354 |
static void |
static void |
1355 |
decoder_bframe(DECODER * dec, |
decoder_bframe(DECODER * dec, |
1356 |
Bitstream * bs, |
Bitstream * bs, |
1362 |
VECTOR mv; |
VECTOR mv; |
1363 |
const VECTOR zeromv = {0,0}; |
const VECTOR zeromv = {0,0}; |
1364 |
int i; |
int i; |
1365 |
|
int resync_len; |
1366 |
|
|
1367 |
if (!dec->is_edged[0]) { |
if (!dec->is_edged[0]) { |
1368 |
start_timer(); |
start_timer(); |
1380 |
stop_edges_timer(); |
stop_edges_timer(); |
1381 |
} |
} |
1382 |
|
|
1383 |
|
resync_len = get_resync_len_b(fcode_backward, fcode_forward); |
1384 |
for (y = 0; y < dec->mb_height; y++) { |
for (y = 0; y < dec->mb_height; y++) { |
1385 |
/* Initialize Pred Motion Vector */ |
/* Initialize Pred Motion Vector */ |
1386 |
dec->p_fmv = dec->p_bmv = zeromv; |
dec->p_fmv = dec->p_bmv = zeromv; |
1387 |
for (x = 0; x < dec->mb_width; x++) { |
for (x = 0; x < dec->mb_width; x++) { |
1388 |
MACROBLOCK *mb = &dec->mbs[y * dec->mb_width + x]; |
MACROBLOCK *mb = &dec->mbs[y * dec->mb_width + x]; |
1389 |
MACROBLOCK *last_mb = &dec->last_mbs[y * dec->mb_width + x]; |
MACROBLOCK *last_mb = &dec->last_mbs[y * dec->mb_width + x]; |
1390 |
const int fcode_max = (fcode_forward>fcode_backward) ? fcode_forward : fcode_backward; |
int intra_dc_threshold; /* fake variable */ |
|
uint32_t intra_dc_threshold; /* fake variable */ |
|
|
|
|
|
if (check_resync_marker(bs, fcode_max - 1)) { |
|
|
int bound = read_video_packet_header(bs, dec, fcode_max - 1, &quant, |
|
|
&fcode_forward, &fcode_backward, &intra_dc_threshold); |
|
|
x = bound % dec->mb_width; |
|
|
y = bound / dec->mb_width; |
|
|
/* reset predicted macroblocks */ |
|
|
dec->p_fmv = dec->p_bmv = zeromv; |
|
|
} |
|
1391 |
|
|
1392 |
mv = |
mv = |
1393 |
mb->b_mvs[0] = mb->b_mvs[1] = mb->b_mvs[2] = mb->b_mvs[3] = |
mb->b_mvs[0] = mb->b_mvs[1] = mb->b_mvs[2] = mb->b_mvs[3] = |
1403 |
if (last_mb->mode == MODE_NOT_CODED) { |
if (last_mb->mode == MODE_NOT_CODED) { |
1404 |
mb->cbp = 0; |
mb->cbp = 0; |
1405 |
mb->mode = MODE_FORWARD; |
mb->mode = MODE_FORWARD; |
1406 |
decoder_mbinter(dec, mb, x, y, mb->cbp, bs, 0, 1); |
decoder_mbinter(dec, mb, x, y, mb->cbp, bs, 0, 1, 1); |
1407 |
continue; |
continue; |
1408 |
} |
} |
1409 |
|
|
1410 |
|
if (check_resync_marker(bs, resync_len)) { |
1411 |
|
int bound = read_video_packet_header(bs, dec, resync_len, &quant, |
1412 |
|
&fcode_forward, &fcode_backward, &intra_dc_threshold); |
1413 |
|
|
1414 |
|
bound = MAX(0, bound-1); /* valid bound must always be >0 */ |
1415 |
|
x = bound % dec->mb_width; |
1416 |
|
y = MIN((bound / dec->mb_width), (dec->mb_height-1)); |
1417 |
|
/* reset predicted macroblocks */ |
1418 |
|
dec->p_fmv = dec->p_bmv = zeromv; |
1419 |
|
/* update resync len with new fcodes */ |
1420 |
|
resync_len = get_resync_len_b(fcode_backward, fcode_forward); |
1421 |
|
continue; /* re-init loop */ |
1422 |
|
} |
1423 |
|
|
1424 |
if (!BitstreamGetBit(bs)) { /* modb=='0' */ |
if (!BitstreamGetBit(bs)) { /* modb=='0' */ |
1425 |
const uint8_t modb2 = BitstreamGetBit(bs); |
const uint8_t modb2 = BitstreamGetBit(bs); |
1426 |
|
|
1500 |
get_b_motion_vector(bs, &mb->mvs[0], fcode_backward, dec->p_bmv, dec, x, y); |
get_b_motion_vector(bs, &mb->mvs[0], fcode_backward, dec->p_bmv, dec, x, y); |
1501 |
dec->p_bmv = mb->mvs[1] = mb->mvs[2] = mb->mvs[3] = mb->mvs[0]; |
dec->p_bmv = mb->mvs[1] = mb->mvs[2] = mb->mvs[3] = mb->mvs[0]; |
1502 |
|
|
1503 |
decoder_mbinter(dec, mb, x, y, mb->cbp, bs, 0, 0); |
decoder_mbinter(dec, mb, x, y, mb->cbp, bs, 0, 0, 1); |
1504 |
break; |
break; |
1505 |
|
|
1506 |
case MODE_FORWARD: |
case MODE_FORWARD: |
1507 |
get_b_motion_vector(bs, &mb->mvs[0], fcode_forward, dec->p_fmv, dec, x, y); |
get_b_motion_vector(bs, &mb->mvs[0], fcode_forward, dec->p_fmv, dec, x, y); |
1508 |
dec->p_fmv = mb->mvs[1] = mb->mvs[2] = mb->mvs[3] = mb->mvs[0]; |
dec->p_fmv = mb->mvs[1] = mb->mvs[2] = mb->mvs[3] = mb->mvs[0]; |
1509 |
|
|
1510 |
decoder_mbinter(dec, mb, x, y, mb->cbp, bs, 0, 1); |
decoder_mbinter(dec, mb, x, y, mb->cbp, bs, 0, 1, 1); |
1511 |
break; |
break; |
1512 |
|
|
1513 |
default: |
default: |
1518 |
} |
} |
1519 |
|
|
1520 |
/* perform post processing if necessary, and output the image */ |
/* perform post processing if necessary, and output the image */ |
1521 |
void decoder_output(DECODER * dec, IMAGE * img, MACROBLOCK * mbs, |
static void decoder_output(DECODER * dec, IMAGE * img, MACROBLOCK * mbs, |
1522 |
xvid_dec_frame_t * frame, xvid_dec_stats_t * stats, |
xvid_dec_frame_t * frame, xvid_dec_stats_t * stats, |
1523 |
int coding_type, int quant) |
int coding_type, int quant) |
1524 |
{ |
{ |
1534 |
image_copy(&dec->tmp, img, dec->edged_width, dec->height); |
image_copy(&dec->tmp, img, dec->edged_width, dec->height); |
1535 |
image_postproc(&dec->postproc, &dec->tmp, dec->edged_width, |
image_postproc(&dec->postproc, &dec->tmp, dec->edged_width, |
1536 |
mbs, dec->mb_width, dec->mb_height, dec->mb_width, |
mbs, dec->mb_width, dec->mb_height, dec->mb_width, |
1537 |
frame->general, brightness, dec->frames, (coding_type == B_VOP)); |
frame->general, brightness, dec->frames, (coding_type == B_VOP), dec->num_threads); |
1538 |
img = &dec->tmp; |
img = &dec->tmp; |
1539 |
} |
} |
1540 |
|
|
1541 |
|
if ((frame->output.csp == XVID_CSP_INTERNAL) || |
1542 |
|
((frame->output.plane[0] != NULL) && (frame->output.stride[0] >= dec->width))) { |
1543 |
image_output(img, dec->width, dec->height, |
image_output(img, dec->width, dec->height, |
1544 |
dec->edged_width, (uint8_t**)frame->output.plane, frame->output.stride, |
dec->edged_width, (uint8_t**)frame->output.plane, frame->output.stride, |
1545 |
frame->output.csp, dec->interlacing); |
frame->output.csp, dec->interlacing); |
1546 |
|
} |
1547 |
|
|
1548 |
if (stats) { |
if (stats) { |
1549 |
stats->type = coding2type(coding_type); |
stats->type = coding2type(coding_type); |
1552 |
stats->data.vop.qscale_stride = dec->mb_width; |
stats->data.vop.qscale_stride = dec->mb_width; |
1553 |
stats->data.vop.qscale = dec->qscale; |
stats->data.vop.qscale = dec->qscale; |
1554 |
if (stats->data.vop.qscale != NULL && mbs != NULL) { |
if (stats->data.vop.qscale != NULL && mbs != NULL) { |
1555 |
int i; |
unsigned int i; |
1556 |
for (i = 0; i < dec->mb_width*dec->mb_height; i++) |
for (i = 0; i < dec->mb_width*dec->mb_height; i++) |
1557 |
stats->data.vop.qscale[i] = mbs[i].quant; |
stats->data.vop.qscale[i] = mbs[i].quant; |
1558 |
} else |
} else |
1566 |
{ |
{ |
1567 |
|
|
1568 |
Bitstream bs; |
Bitstream bs; |
1569 |
uint32_t rounding; |
uint32_t rounding = 0; |
1570 |
uint32_t quant = 2; |
uint32_t quant = 2; |
1571 |
uint32_t fcode_forward; |
uint32_t fcode_forward = 0; |
1572 |
uint32_t fcode_backward; |
uint32_t fcode_backward = 0; |
1573 |
uint32_t intra_dc_threshold; |
uint32_t intra_dc_threshold = 0; |
1574 |
WARPPOINTS gmc_warp; |
WARPPOINTS gmc_warp; |
1575 |
int coding_type; |
int coding_type = -1; |
1576 |
int success, output, seen_something; |
int success, output, seen_something; |
1577 |
|
|
1578 |
if (XVID_VERSION_MAJOR(frame->version) != 1 || (stats && XVID_VERSION_MAJOR(stats->version) != 1)) /* v1.x.x */ |
if (XVID_VERSION_MAJOR(frame->version) != 1 || (stats && XVID_VERSION_MAJOR(stats->version) != 1)) /* v1.x.x */ |
1579 |
return XVID_ERR_VERSION; |
return XVID_ERR_VERSION; |
1580 |
|
|
1581 |
start_global_timer(); |
start_global_timer(); |
1582 |
|
memset((void *)&gmc_warp, 0, sizeof(WARPPOINTS)); |
1583 |
|
|
1584 |
dec->low_delay_default = (frame->general & XVID_LOWDELAY); |
dec->low_delay_default = (frame->general & XVID_LOWDELAY); |
1585 |
if ((frame->general & XVID_DISCONTINUITY)) |
if ((frame->general & XVID_DISCONTINUITY)) |
1644 |
if (coding_type == -2 || coding_type == -3) { /* vol and/or resize */ |
if (coding_type == -2 || coding_type == -3) { /* vol and/or resize */ |
1645 |
|
|
1646 |
if (coding_type == -3) |
if (coding_type == -3) |
1647 |
decoder_resize(dec); |
if (decoder_resize(dec)) return XVID_ERR_MEMORY; |
1648 |
|
|
1649 |
if (stats) { |
if (stats) { |
1650 |
stats->type = XVID_TYPE_VOL; |
stats->type = XVID_TYPE_VOL; |
1651 |
stats->data.vol.general = 0; |
stats->data.vol.general = 0; |
1652 |
/*XXX: if (dec->interlacing) |
stats->data.vop.general = 0; |
1653 |
stats->data.vol.general |= ++INTERLACING; */ |
if (dec->interlacing) { |
1654 |
|
stats->data.vol.general |= XVID_VOL_INTERLACING; |
1655 |
|
if (dec->top_field_first) { |
1656 |
|
stats->data.vop.general |= XVID_VOP_TOPFIELDFIRST; |
1657 |
|
} |
1658 |
|
} |
1659 |
stats->data.vol.width = dec->width; |
stats->data.vol.width = dec->width; |
1660 |
stats->data.vol.height = dec->height; |
stats->data.vol.height = dec->height; |
1661 |
stats->data.vol.par = dec->aspect_ratio; |
stats->data.vol.par = dec->aspect_ratio; |
1667 |
goto repeat; |
goto repeat; |
1668 |
} |
} |
1669 |
|
|
1670 |
if(dec->frames == 0 && coding_type != I_VOP) { |
if((dec->frames == 0 && coding_type != I_VOP) || (!dec->width || !dec->height)) { |
1671 |
/* 1st frame is not an i-vop */ |
/* 1st frame is not an i-vop */ |
1672 |
goto repeat; |
goto repeat; |
1673 |
} |
} |
1674 |
|
|
1675 |
dec->p_bmv.x = dec->p_bmv.y = dec->p_fmv.y = dec->p_fmv.y = 0; /* init pred vector to 0 */ |
dec->p_bmv.x = dec->p_bmv.y = dec->p_fmv.x = dec->p_fmv.y = 0; /* init pred vector to 0 */ |
1676 |
|
|
1677 |
/* packed_mode: special-N_VOP treament */ |
/* packed_mode: special-N_VOP treament */ |
1678 |
if (dec->packed_mode && coding_type == N_VOP) { |
if (dec->packed_mode && coding_type == N_VOP) { |